A laser positioning method and system for port yard transfer

Abstract

The application discloses a laser positioning method and system for port yard transfer, classifies key environmental elements in a port yard, performs semantic segmentation after constructing a point cloud map of the port yard, retains long-term static targets, removes dynamic targets, extracts short-term static targets, constructs a short-term static target point cloud template in combination with actual physical sizes of the short-term static targets, and performs filling and other processing on corresponding positions of the short-term static targets in the point cloud map, so that positioning failure caused by insufficient available features during matching positioning is avoided, in addition, the point cloud map is subjected to rasterization processing, the influence of point cloud distribution in the template on original point cloud is reduced, a processed positioning layer is obtained, and in combination with matching positioning, stable positioning results can be output when the yard environment changes greatly, and the positioning performance of a port vehicle is improved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent driving technology and relates to the positioning of port yard transfer vehicles, specifically to a laser positioning method and system for port yard transfer. Background Technology

[0002] With the rapid development of my country's maritime transport, port terminals, as the most important transportation hubs, have seen an average annual container throughput exceeding 200 million TEUs in recent years, making efficiency particularly crucial. For cargo loading and unloading, port terminals mainly employ three types of equipment: quay cranes for loading and unloading ships at the quay, horizontal transport equipment moving between the quay and the container yard, and yard cranes for loading and unloading containers within the container yard. Currently, horizontal transport is primarily handled by driver-operated container trucks, and efficiency and labor costs are key challenges that port terminals urgently need to address.

[0003] Currently, the commonly used technical solutions include two types: one is the automated guided vehicle (AGV) with magnetic nail navigation, and the other is the autonomous vehicle equipped with relevant perception and positioning sensors. The AGV with magnetic nail navigation replaces the manned container truck, but it requires burying a large number of magnetic nails in the ground, and the site modification and subsequent maintenance costs are very high. In comparison, the autonomous vehicle equipped with relevant perception and positioning sensors is obviously a more effective way to comprehensively improve safety and efficiency.

[0004] For autonomous vehicles, vehicle localization is a key technology. However, in port yards, due to the height of stacked objects, GPS signal obstruction is often severe, resulting in poor real-time differential positioning performance. Currently, port autonomous driving relies heavily on maps; therefore, map-based relative positioning is a common auxiliary positioning method. Current visual positioning solutions largely depend on site modifications and struggle to achieve stable results in areas where ground markings are easily worn away. In contrast, laser positioning, with its high ranging accuracy and immunity to ambient light, is a relatively more stable solution.

[0005] For scenarios like port yards where the environment changes frequently, the industry often uses the method of periodically updating the positioning layer. However, this method has problems such as long update cycles and high maintenance costs. At the same time, multiple map updates inevitably introduce cumulative errors. These factors pose a great challenge to the effectiveness of map-based laser positioning. Therefore, solving the problem of stable positioning in port yard environments has become an urgent issue. Summary of the Invention

[0006] To address the aforementioned problems, the main objective of this invention is to design a laser positioning method and system for port yard transshipment, thereby solving the positioning stability issue for port vehicles.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: A laser positioning method for transshipment in port yards, comprising generating a positioning layer for matching and matching positioning based on the positioning layer, as detailed below: Step 1: Collect mapping data and construct an original point cloud map based on the collected mapping data; Step 2: Perform semantic segmentation on the original point cloud map to identify dynamic targets and short-term static targets. Remove the point cloud corresponding to the dynamic targets in the original point cloud map to obtain the base map, and construct a short-term static target point cloud template based on the base map. Step 3: Based on the short-term static target point cloud template, perform post-processing on the specified area of ​​the base map to obtain a complete point cloud map; Step 4: Rasterize the complete point cloud map to obtain the positioning layer for matching; Step 5: Acquire sensor data in real time and process the format of the sensor data; Step 6: Preprocess the processed data to obtain a local point cloud map; Step 7: Load the positioning layer obtained in Step 4 for matching, match the local point cloud map with it, and obtain the vehicle pose. Step 8: Output the vehicle positioning results.

[0008] As a further description of the present invention, after the short-term static target point cloud template constructed in step 2, another setting is included, specifically: Step 3: Perform rasterization processing on the base map and the short-term static target point cloud template respectively; Step 4: Based on the short-term static target point cloud template after rasterization, perform post-processing on the specified area of ​​the rasterized base map to obtain the positioning layer used for matching.

[0009] As a further description of the present invention, in step 1, an original point cloud map is constructed using the collected mapping data through a high-precision integrated navigation system and real-time positioning and mapping method.

[0010] As a further description of the present invention, the short-term static target includes one, more or all of the following: containers, prefabricated houses, road bollards, and cones; The dynamic targets mentioned include one, more, or all of the following: vehicles, people, and cranes.

[0011] As a further description of the present invention, in step 2, the point cloud corresponding to the short-term static target in the obtained base map is reconstructed in three dimensions to obtain a short-term static target point cloud template.

[0012] As a further description of the present invention, the designated area is the possible placement location of the short-term static target in the base map; Post-processing involves incremental filling of the specified area.

[0013] As a further description of the present invention, the specific resolution of the rasterization process depends on the viewpoint cloud map and the sensor accuracy, and the raster form includes occupancy raster and probability raster.

[0014] As a further description of the present invention, in step 6, the preprocessing of point cloud data includes one, more or all of the following: point cloud stitching, point cloud distortion removal, and downsampling.

[0015] As a further description of the present invention, in step 7, the positioning layer used for matching is matched with the local point cloud map, and the matching method is to obtain the vehicle pose with the highest matching score. The matching score is the sum of the probabilities that all points in the local point cloud map fall within the corresponding grid cells of the positioning layer used for matching.

[0016] As a further description of the present invention, in step 8, the vehicle matching result is converted into a data format that conforms to the communication protocol and then output.

[0017] A laser positioning system for transshipment in port yards includes a point cloud template construction module, a positioning layer processing module, a matching positioning module, and a positioning result output module. The system's inputs include point cloud map data needed to construct the positioning layer and sensor data required for online operation; The point cloud map data required for constructing the positioning layer is transmitted to the positioning layer processing module. The point cloud template construction module and the positioning layer processing module are configured to be mutually transferable, and the positioning layer processing module is connected to the matching positioning module. The required sensor data is transmitted online to the matching and positioning module, and the output of the matching and positioning module is connected to the positioning result output module.

[0018] As a further description of the present invention, the system also includes: a sensor data acquisition module for real-time acquisition of vehicle data and a point cloud preprocessing module for processing the acquired data; The sensor data required for online processing is acquired by the sensor data acquisition module, and then the acquired sensor data is transmitted to the point cloud preprocessing module, which is connected to the matching and positioning module.

[0019] Compared with the prior art, the technical advantages of the present invention are as follows: This invention provides a laser positioning method and system for port yard transshipment. Key environmental elements in the port yard are categorized based on their frequency of change, roughly into long-term static targets, short-term static targets, and dynamic targets. After constructing a point cloud map of the port yard, semantic segmentation is performed, retaining long-term static targets and removing dynamic targets. Short-term static targets are then extracted and their actual physical dimensions are combined to construct a short-term static target point cloud template. Filling is performed on the corresponding positions of the short-term static targets in the point cloud map to avoid positioning failures due to insufficient available features during matching. Furthermore, the point cloud map is rasterized to reduce the impact of inconsistencies between the point cloud distribution in the template and the original point cloud, resulting in a processed positioning layer.

[0020] Meanwhile, by constructing short-term static target point cloud templates for port yards, we can ensure that we obtain fresh and complete positioning layers for the port yards, reducing later maintenance costs. Combined with matching positioning, we can ensure that stable positioning results can still be output when the yard environment changes significantly, thus improving the positioning performance of port vehicles. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the positioning layer generation process of the present invention; Figure 2 This is a schematic diagram of the matching and positioning process in this invention; Figure 3 This is a schematic diagram of the point cloud filling effect in this invention; Figure 4 This is a schematic diagram showing the coordination of the various modules of the positioning system of the present invention. Detailed Implementation

[0022] The present invention will now be described in detail with reference to the accompanying drawings: A laser positioning method for transshipment in port yards, reference Figure 1-2 As shown, the method includes generating a positioning layer for matching and matching positioning based on the positioning layer, as detailed below: Step 1: Collect mapping data and construct an original point cloud map based on the collected mapping data; Step 2: Perform semantic segmentation on the original point cloud map to identify dynamic targets and short-term static targets. Remove the point cloud corresponding to the dynamic targets in the original point cloud map to obtain the base map, and construct a short-term static target point cloud template based on the base map. Step 3: Based on the short-term static target point cloud template, perform post-processing on the specified area of ​​the base map to obtain a complete point cloud map; Step 4: Rasterize the complete point cloud map to obtain the positioning layer for matching; Step 5: Acquire sensor data in real time and process the format of the sensor data; Step 6: Preprocess the processed data to obtain a local point cloud map; Step 7: Load the positioning layer obtained in Step 4 for matching, match the local point cloud map with it, and obtain the vehicle pose. Step 8: Output the vehicle positioning results.

[0023] In another embodiment of the present invention, a laser positioning method for transshipment in a port yard includes generating a positioning layer for matching and matching positioning based on the positioning layer, as detailed below: Step 1: Collect mapping data and construct an original point cloud map based on the collected mapping data; Step 2: Perform semantic segmentation on the original point cloud map to identify dynamic targets and short-term static targets. Remove the point cloud corresponding to the dynamic targets in the original point cloud map to obtain the base map, and construct a short-term static target point cloud template based on the base map. Step 3: Perform rasterization processing on the base map and the short-term static target point cloud template respectively; Step 4: Based on the rasterized short-term static target point cloud template, perform post-processing on the specified area of ​​the rasterized base map to obtain the positioning layer used for matching; Step 5: Acquire sensor data in real time and process the format of the sensor data; Step 6: Preprocess the processed data to obtain a local point cloud map; Step 7: Load the positioning layer obtained in Step 4 for matching, match the local point cloud map with it, and obtain the vehicle pose. Step 8: Output the vehicle positioning results.

[0024] Specifically, the overall positioning method of this invention includes generating a positioning layer for matching and matching positioning based on the positioning layer; The process of generating the positioning layer for matching is as follows: First, key environmental elements in the port yard are classified, roughly into long-term static targets, short-term static targets, and dynamic targets based on the frequency of changes. After constructing the point cloud map of the port yard, semantic segmentation is performed, retaining long-term static targets and removing dynamic targets. Then, short-term static targets are extracted and combined with their actual physical dimensions to construct a short-term static target point cloud template. Next, filling in the corresponding positions of the short-term static targets in the point cloud map is performed to avoid positioning failures due to insufficient available features during matching. Finally, the point cloud map is rasterized to reduce the impact of inconsistencies between the point cloud distribution in the template and the original point cloud, resulting in a processed positioning layer, such as... Figure 1 As shown.

[0025] The matching and localization process based on the localization layer is as follows: First, the point clouds scanned by one or more LiDAR sensors are preprocessed, including stitching, motion compensation, and downsampling. Then, a grid-based matching method is used to match the processed local point cloud map with the localization layer to obtain the vehicle's pose relative to the map. Figure 2 As shown.

[0026] This invention provides a detailed analysis of the implementation methods of the two positioning methods described above, as follows: I. Methods for generating and processing positioning layers, including: a. Construct point cloud templates for short-term static targets in the yard environment; b. Perform filling and other processing on the specified area of ​​the point cloud map according to the point cloud template to ensure the freshness and integrity of the point cloud map; c. Rasterize the point cloud map to reduce the impact of inconsistencies between the point cloud portion in the template and the original point cloud, as well as the physical memory required to fill in the large amount of point cloud.

[0027] The main process steps are as follows: (1) Collect port yard mapping data; (2) Using the collected mapping data, construct the original point cloud map of the port yard; (3) Perform semantic segmentation on the original point cloud map to identify dynamic targets and short-term static targets, and remove the point cloud corresponding to the dynamic targets in the point cloud map to obtain the base map; (4) Construct a short-term static target point cloud template; (5) In the specified area of ​​the map (referring to the possible placement of short-term static targets in the map), post-process (point cloud filling, etc.) using the short-term static target point cloud template to obtain a fresh and complete point cloud map; (6) Rasterize the point cloud map to obtain a layer that can be used for matching and positioning.

[0028] More specifically, the mapping data in step (1) can be obtained by means of, but not limited to, map collection vehicles, port autonomous vehicles, and handheld mapping devices.

[0029] Step (2) employs, but is not limited to, high-precision integrated navigation systems, real-time positioning and mapping technologies, to construct the original point cloud map of the storage yard using the collected mapping data; It should be noted that since step (2) stitches together all the collected point clouds based on the positioning information of the mapping equipment, the constructed point cloud map... Figure 1On the one hand, there are a large number of dynamic targets such as people and vehicles, which will interfere with the matching and positioning; on the other hand, there are a large number of short-term static targets such as containers in the port yard, and there are some empty container areas in the constructed point cloud map, which causes the positioning to fail due to the lack of sufficient matching features. Therefore, steps (3), (4), and (5) are still required.

[0030] Step (3) can be performed by training a deep learning model or by interactively segmenting dynamic targets and short-term static targets from a point cloud map using visualization software. Point cloud visualization software includes, but is not limited to, CloudCompare and MeshLab. Similarly, the point cloud corresponding to dynamic targets can be removed by methods including, but not limited to, categories detected by deep learning models and selection by visualization software. Dynamic targets include, but are not limited to, vehicles, people, and tire cranes.

[0031] The short-term static targets in step (4) include, but are not limited to, containers, prefabricated houses, road bollards, and traffic cones; Optional point cloud template generation methods include, but are not limited to, directly using the semantic segmentation results of point cloud maps, separately collecting the corresponding target point cloud for 3D reconstruction, and modeling based on the corresponding CAD drawings.

[0032] Step (5) can be based directly on the location of existing short-term static targets in the point cloud map, or it can be combined with vector map information to incrementally fill the point cloud template at a specified location in the point cloud map; Specifically, taking the point cloud filling process of containers as an example, equidistant filling can be performed in the point cloud map by combining the existing container point cloud distribution in the stacking area and the actual placement rules. Alternatively, vector information of container positions in the map can be extracted for point cloud filling. The filling effect is as follows: Figure 3 As shown. Similarly, when the state of long-term goals changes, the same method can be used to quickly update the map, reducing map maintenance costs.

[0033] Step (6) divides the map into grids according to a certain resolution to reduce the impact of inconsistencies between the point cloud distribution in the template and the original point cloud. The resolution can be determined based on the accuracy of the map and the sensor; the higher the accuracy, the smaller the map resolution can be set. Grids can include, but are not limited to, occupied grids and probabilistic grids. The probability model in the probabilistic grid can adopt, but is not limited to, Gaussian distribution, complex Gaussian distribution, and Gaussian mixture distribution.

[0034] II. Matching and positioning methods based on positioning layers, including: a. Acquire sensor data from the vehicle in real time and process the data format required for positioning; b. Preprocess the point cloud data in the processing format to obtain a local point cloud map; c. Using a grid-based matching method, the local point cloud map is matched with the local positioning layer used for matching to obtain the vehicle's pose relative to the local positioning layer used for matching.

[0035] The main process steps are as follows: (1) Acquire sensor data in real time and process it into the data format required for positioning; (2) Perform preprocessing such as stitching, motion compensation, and downsampling on the point cloud data of the processing format to obtain a local point cloud map; (3) Load the positioning layer for matching, and match the local point cloud map with the positioning layer to obtain the vehicle's pose; (4) Output the vehicle positioning results.

[0036] More specifically, the sensor data in step (1) includes, but is not limited to, laser point cloud, IMU, and wheel speed data. Among them, the laser point cloud data can be acquired by a single radar or multiple radars, and the types of lidar devices include, but are not limited to, mechanical rotating type and MEMS type.

[0037] It should be noted that the processing method of sensor data depends on the output format of the original sensor and the data structure defined inside the positioning program. Taking the commonly used general format as an example, the output of LiDAR sensors is often in pcap (PacketCapture Data) format, and the commonly used processed data is in pcl (Point Cloud Library) point cloud format. A corresponding driver module needs to be written to parse the pcap data into pcl point cloud.

[0038] The point cloud preprocessing steps in step (2) include, but are not limited to, point cloud stitching, point cloud distortion correction, and downsampling. The point clouds used in the stitching process include, but are not limited to, multi-frame point clouds acquired by a single lidar or single-frame point clouds acquired by multiple lidars. The distortion correction process can be completed using, but is not limited to, odometry and dead reckoning results. The downsampling process can be completed using, but is not limited to, voxel grid downsampling, uniform downsampling, and random downsampling methods. Through these steps, a modified local point cloud map with relatively rich features can be obtained.

[0039] Step (3) can be matched using optimization methods including but not limited to gradient descent, Gauss-Newton method, and conjugate gradient method. The optimization goal is to obtain a pose solution that makes the local point cloud map output in step (2) match the positioning layer with the highest score. The matching score depends on the probability that all points in the local point cloud map fall into the corresponding grid of the positioning layer.

[0040] Step (4) converts the vehicle matching result into a data format that conforms to the communication protocol and outputs it to downstream modules such as planning and control for vehicle control.

[0041] Another embodiment of the laser positioning method of the present invention involves rasterizing the base map and the short-term static target point cloud template respectively, and then performing post-processing on a designated area of ​​the rasterized base map based on the rasterized short-term static target point cloud template to obtain a positioning layer for matching. Other steps are implemented according to the above content.

[0042] In another embodiment of the invention, a laser positioning system for port yard transshipment is also included, such as... Figure 4 As shown, it is applied to port vehicles and includes a point cloud template construction module, a positioning layer processing module, a matching positioning module, and a positioning result output module; The system's inputs include point cloud map data required to construct the positioning layer and sensor data required for online operation. The point cloud map data required to construct the positioning layer is transmitted to the positioning layer processing module. The point cloud template construction module and the positioning layer processing module are configured to transmit data between each other, and the positioning layer processing module is connected to the matching positioning module. The sensor data required for online operation is transmitted to the matching positioning module, and the output of the matching positioning module is connected to the positioning result output module.

[0043] In addition, the system also includes: a sensor data acquisition module for real-time vehicle data acquisition and a point cloud preprocessing module for processing the acquired data; the sensor data required online is acquired through the sensor data acquisition module, and then the acquired sensor data is transmitted to the point cloud preprocessing module, which is connected to the matching and positioning module.

[0044] More specifically, in this embodiment, the functions of each module are analyzed in detail, and the analysis content is as follows: Sensor data acquisition module: used to acquire sensor data on the vehicle in real time and convert it into the data format required for positioning; Point cloud preprocessing module: used to perform preprocessing such as stitching, motion compensation, and downsampling on the acquired laser point cloud data to obtain a local point cloud map; Point cloud template building module: used to generate short-term static target point cloud templates for port yards; Location layer processing module: used to process the original point cloud map to obtain a fresh and complete location layer; Matching and positioning module: used to match the local point cloud map with the positioning layer to obtain the vehicle's current pose; Positioning output module: Used to output positioning results that conform to the data communication protocol to downstream planning and control.

[0045] It should also be noted that the specific workflow of the laser positioning system for port yard transshipment disclosed in this embodiment is as follows: The inputs to the entire system include: point cloud map data required to construct the positioning layer and sensor data required for online positioning; the sensor data required for online positioning is acquired through the sensor data acquisition module, and the acquired sensor data is processed by the point cloud preprocessing module to obtain a local point cloud map; S1: The positioning layer processing module performs semantic segmentation on the point cloud map data, classifying the elements into long-term static targets, short-term static targets, and dynamic targets based on the frequency of changes in the elements, retaining long-term static targets and removing dynamic targets. S2: Extract short-term static targets and input them into the point cloud template construction module; S3: The point cloud template construction module combines semantic segmentation to construct point cloud templates from different short-term static target point clouds and their corresponding actual physical size information. S4: Input the constructed point cloud template into the positioning layer processing module; S5: The positioning layer processing module fills the template at the corresponding position of the point cloud base map to obtain a complete point cloud map, and performs rasterization processing to obtain a positioning layer that can be used for positioning, and outputs it to the matching positioning module. S6: The matching and positioning module receives the local point cloud map processed by the point cloud preprocessing module, loads the preprocessed positioning layer, and performs matching and positioning. S7: Outputs positioning results conforming to the data communication protocol through the positioning output module.

[0046] This invention constructs a short-term static target point cloud template for port yards, ensuring that a fresh and complete positioning layer of the port yard can be obtained, reducing subsequent maintenance costs. In addition, the matching method based on the positioning layer ensures that stable positioning results can still be output when the yard environment undergoes significant changes, thereby improving the positioning performance of port vehicles.

[0047] The advantages of this invention compared to existing technologies are analyzed as follows: 1. Existing technologies typically involve pre-positioning markers such as magnetic nails, reflective road studs, or reflective boards in the site area, which requires site-side modifications, resulting in high costs and difficulties in subsequent maintenance. Furthermore, existing technologies mainly rely on visual sensors for positioning, but visual sensors have relatively low ranging accuracy, and in places like ports where ground markers are easily worn, visual positioning is difficult to achieve a stable effect.

[0048] The proposed solution is a laser positioning scheme, which is unaffected by the environment, requires no site modification, and has advantages in stability and cost.

[0049] 2. Currently, laser positioning is relatively widely researched and applied in special scenarios such as quay cranes and cranes in ports. Within quay cranes and container areas, there are solutions that achieve lateral positioning of vehicles by detecting the distance between the vehicle and targets on both sides of the road; however, these only solve the problem of lateral positioning relative to a specific target. There are also solutions that achieve vehicle positioning by detecting geometric features such as lane lines through laser point cloud reflection intensity or other methods. However, these solutions share similar problems with visual positioning solutions, requiring high clarity of roadside markings and advanced sensor and perception performance. Furthermore, optimization schemes have been proposed to address the issues of numerous dynamic targets in ports, sparse low-beam laser radar point clouds, and the speed of point cloud matching pose calculation. Solutions for multi-sensor fusion positioning have also been proposed, but these solutions only improve positioning stability to a certain extent when the scene remains unchanged; they still do not specifically address the positioning problems caused by frequent scene changes during container yard transfers.

[0050] This invention ensures the integrity and freshness of the positioning layer by constructing a short-term static target template and updating and completing it at designated locations on the port yard point cloud base map. Compared with existing solutions, this method has a shorter update cycle and lower maintenance costs. Vehicle pose is then obtained through matching and positioning based on this layer, specifically addressing the matching and positioning failure problem caused by significant changes in port yard transshipment scenarios.

[0051] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Any other modifications or equivalent substitutions made by those skilled in the art to the technical solutions of the present invention, as long as they do not depart from the spirit and scope of the technical solutions of the present invention, should be covered within the scope of the claims of the present invention.

Claims

1. A laser positioning method for transshipment in port yards, characterized in that: This method includes generating a positioning layer for matching and matching positioning based on the positioning layer, as follows: Step 1: Collect mapping data and construct an original point cloud map based on the collected mapping data; Step 2: Perform semantic segmentation on the original point cloud map to identify dynamic targets and short-term static targets. Remove the point cloud corresponding to the dynamic targets in the original point cloud map to obtain the base map, and construct a short-term static target point cloud template based on the base map. The short-term static targets include one, more, or all of the following: containers, prefabricated houses, road bollards, and traffic cones; the dynamic targets include one, more, or all of the following: vehicles, people, and cranes. Step 3: Based on the short-term static target point cloud template, perform post-processing on the specified area of ​​the base map to obtain a complete point cloud map; Step 4: Rasterize the complete point cloud map to obtain the positioning layer for matching; Step 5: Acquire sensor data in real time and process the format of the sensor data; Step 6: Preprocess the processed data to obtain a local point cloud map; Step 7: Load the positioning layer obtained in Step 4 for matching, match the local point cloud map with it, and obtain the vehicle pose. Step 8: Output the vehicle positioning results.

2. The laser positioning method for port yard transshipment according to claim 1, characterized in that: Following the short-term static target point cloud template constructed in step 2, another setting is included, specifically: Step 3: Perform rasterization processing on the base map and the short-term static target point cloud template respectively; Step 4: Based on the short-term static target point cloud template after rasterization, perform post-processing on the specified area of ​​the rasterized base map to obtain the positioning layer used for matching.

3. The laser positioning method for port yard transshipment according to claim 2, characterized in that: In step 2, the point cloud corresponding to the short-term static target in the obtained base map is reconstructed in three dimensions to obtain the short-term static target point cloud template.

4. A laser positioning method for port yard transshipment according to claim 1 or 2, characterized in that: The designated area refers to the possible placement of short-term static targets on the base map; Post-processing involves incremental filling of the specified area.

5. A laser positioning method for port yard transshipment according to claim 1 or 2, characterized in that: The specific resolution of the rasterization process depends on the viewpoint cloud map and the sensor accuracy. The raster types include occupancy raster and probabilistic raster.

6. A laser positioning method for port yard transshipment according to claim 1 or 2, characterized in that: In step 7, the positioning layer used for matching is matched with the local point cloud map, and the matching method is to obtain the vehicle pose with the highest matching score. The matching score is the sum of the probabilities that all points in the local point cloud map fall within the corresponding grid cells of the positioning layer used for matching.

7. A laser positioning method for port yard transshipment according to claim 1 or 2, characterized in that: In step 8, the vehicle matching results are converted into a data format that conforms to the communication protocol and then output.

8. A laser positioning system for transshipment in port yards, characterized in that: The positioning system is used to perform the positioning method according to any one of claims 1-7, and includes a point cloud template construction module, a positioning layer processing module, a matching positioning module, and a positioning result output module; The system's inputs include point cloud map data needed to construct the positioning layer and sensor data required for online operation; The point cloud map data required for constructing the positioning layer is transmitted to the positioning layer processing module. The point cloud template construction module and the positioning layer processing module are configured to be mutually transferable, and the positioning layer processing module is connected to the matching positioning module. The required sensor data is transmitted online to the matching and positioning module, and the output of the matching and positioning module is connected to the positioning result output module.

9. A laser positioning system for port yard transshipment according to claim 8, characterized in that: The system also includes: a sensor data acquisition module for real-time vehicle data acquisition and a point cloud preprocessing module for processing the acquired data; The sensor data required for online processing is acquired by the sensor data acquisition module, and then the acquired sensor data is transmitted to the point cloud preprocessing module, which is connected to the matching and positioning module.

Images